Skip to content

[BugFix] Fix tpu_model_runner block_id concatenation #19228

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Jun 6, 2025
Merged

[BugFix] Fix tpu_model_runner block_id concatenation #19228

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
1c33ebf
[BugFix] Fix tpu_model_runner block_id concatenation
njhill Jun 5, 2025
bffbd54
also fix test
njhill Jun 6, 2025
53cc24d
Merge remote-tracking branch 'origin/main' into fix-tpu-preempt
njhill Jun 6, 2025
4c721ad
Merge remote-tracking branch 'origin/main' into fix-tpu-preempt
njhill Jun 6, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion tests/v1/tpu/worker/test_tpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -226,7 +226,7 @@ def test_update_states_request_resumed(model_runner):
req_id=req_id,
resumed_from_preemption=False,
new_token_ids=[],
new_block_ids=[],
new_block_ids=[[]],
num_computed_tokens=0,
)

Expand Down
7 changes: 5 additions & 2 deletions vllm/v1/worker/gpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -460,8 +460,11 @@ def _update_states(self, scheduler_output: "SchedulerOutput") -> None:
# Update the block IDs.
if not req_data.resumed_from_preemption:
# Append the new blocks to the existing block IDs.
for i in range(len(self.kv_cache_config.kv_cache_groups)):
req_state.block_ids[i].extend(req_data.new_block_ids[i])
for block_ids, new_block_ids in zip( # type: ignore[call-overload]
req_state.block_ids,
req_data.new_block_ids,
strict=True):
block_ids.extend(new_block_ids)
else:
# The request is resumed from preemption.
# Replace the existing block IDs with the new ones.
Expand Down
6 changes: 5 additions & 1 deletion vllm/v1/worker/tpu_model_runner.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -413,7 +413,11 @@ def _update_states(self, scheduler_output: "SchedulerOutput") -> bool:
req_state.num_computed_tokens = req_data.num_computed_tokens
if not req_data.resumed_from_preemption:
# Append the new blocks to the existing block IDs.
req_state.block_ids.extend(req_data.new_block_ids)
for block_ids, new_block_ids in zip( # type: ignore[call-overload]
req_state.block_ids,
req_data.new_block_ids,
strict=True):
block_ids.extend(new_block_ids)
else:
# The request is resumed from preemption.
# Replace the existing block IDs with the new ones.
Expand Down